CN113965362B - Internet of things intelligent protection system and method based on WiFi signal sensing - Google Patents
Internet of things intelligent protection system and method based on WiFi signal sensing Download PDFInfo
- Publication number
- CN113965362B CN113965362B CN202111183088.1A CN202111183088A CN113965362B CN 113965362 B CN113965362 B CN 113965362B CN 202111183088 A CN202111183088 A CN 202111183088A CN 113965362 B CN113965362 B CN 113965362B
- Authority
- CN
- China
- Prior art keywords
- wifi
- data
- csi
- monitoring
- behaviors
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 32
- 206010000117 Abnormal behaviour Diseases 0.000 claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000013135 deep learning Methods 0.000 claims abstract description 14
- 230000006399 behavior Effects 0.000 claims description 67
- 238000012544 monitoring process Methods 0.000 claims description 49
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 27
- 230000006870 function Effects 0.000 claims description 19
- 239000011159 matrix material Substances 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 15
- 230000008569 process Effects 0.000 claims description 13
- 230000004913 activation Effects 0.000 claims description 12
- 230000000875 corresponding effect Effects 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 9
- 230000004044 response Effects 0.000 claims description 9
- 230000009466 transformation Effects 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 7
- 238000000354 decomposition reaction Methods 0.000 claims description 6
- 238000010200 validation analysis Methods 0.000 claims description 6
- 230000000007 visual effect Effects 0.000 claims description 6
- 230000002159 abnormal effect Effects 0.000 claims description 5
- 238000000513 principal component analysis Methods 0.000 claims description 5
- 238000011161 development Methods 0.000 claims description 4
- 230000018109 developmental process Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 238000011176 pooling Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000013522 software testing Methods 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims 1
- 238000000638 solvent extraction Methods 0.000 claims 1
- 238000001514 detection method Methods 0.000 abstract description 9
- 238000009472 formulation Methods 0.000 abstract 1
- 239000000203 mixture Substances 0.000 abstract 1
- 239000000725 suspension Substances 0.000 abstract 1
- 238000005516 engineering process Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 4
- 230000008447 perception Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 3
- 230000007613 environmental effect Effects 0.000 description 3
- 230000009471 action Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000010295 mobile communication Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 230000033772 system development Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1408—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic by monitoring network traffic
- H04L63/1425—Traffic logging, e.g. anomaly detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L41/00—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks
- H04L41/22—Arrangements for maintenance, administration or management of data switching networks, e.g. of packet switching networks comprising specially adapted graphical user interfaces [GUI]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L43/00—Arrangements for monitoring or testing data switching networks
- H04L43/04—Processing captured monitoring data, e.g. for logfile generation
- H04L43/045—Processing captured monitoring data, e.g. for logfile generation for graphical visualisation of monitoring data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/02—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP]
- H04L67/025—Protocols based on web technology, e.g. hypertext transfer protocol [HTTP] for remote control or remote monitoring of applications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
- H04L67/125—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks involving control of end-device applications over a network
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/02—Total factory control, e.g. smart factories, flexible manufacturing systems [FMS] or integrated manufacturing systems [IMS]
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computer Security & Cryptography (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Human Computer Interaction (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Alarm Systems (AREA)
Abstract
The invention provides an Internet of things intelligent protection system and method based on WiFi signal sensing, and belongs to the technical field of intrusion detection. According to the method, common WiFi signals in families are utilized to sample CSI in the family, recognition of abnormal behaviors of human targets in the home environment of the Internet of things is achieved through a deep learning algorithm model after training and improvement, a cross-layer intelligent protection system is established in cooperation with a specific abnormal behavior formulation strategy, and recognition and suspension of the abnormal behaviors are achieved. The intelligent home protection system is low in cost, high in precision and good in privacy, solves the problem that the intelligent home system is not humanized enough, and encourages participants to use the intelligent home.
Description
Technical Field
The invention belongs to the technical field of intrusion detection, and relates to an Internet of things home furnishing intelligent protection system and method based on WiFi signal perception.
Background
With the continuous development of mobile communication technology, internet technology and artificial intelligence application, smart homes become the direction and trend of home system development. However, people should also think about the information security hidden danger existing behind these internet of things products while enjoying intelligent convenient services. Since the smart home contains a large number of internet of things devices and needs to access the internet for data exchange, there is a great security risk, and the problems caused thereby are also spread all over the world. For example, a remote attacker may hijack an application and play music over the internet to disturb the user. For these problems, all countries actively explore how to improve the security of smart homes, how to effectively protect user privacy, how to resist the intrusion of external hackers to users, and how to improve the detection accuracy and confidentiality of indoor intrusion detection systems. At present, many intrusion detection systems adopting different devices (such as a camera and an infrared device) are appeared at home and abroad, but the detection systems have the defects of poor privacy, higher deployment cost of special devices and the like. Aiming at the problems, the invention seeks an intrusion detection method by taking common WiFi signals as research objects. The WiFi signal is wide in coverage and is a universal signal, however, the application of the WiFi signal is limited to data communication at the end of the Internet at present, and the potential application value of the WiFi signal is not mined. The WiFi-CSI signal records its attenuation factor on each transmission path, which includes information such as scattering, distance attenuation, etc. The WiFi-CSI signal which is propagated by multipath carries a large amount of environment information on propagation paths, so that the environment information implicit in the signal can be mined from the WiFi-CSI. The invention designs and realizes the Internet of things intelligent protection system with low cost, high precision and good privacy.
Disclosure of Invention
The invention solves the problem of how to use the WiFi-CSI data of the physical layer to realize the judgment and identification of abnormal behaviors through a deep learning algorithm model. According to the invention, passive sensing is taken as research content, a high-precision identification system based on deep learning is realized, technologies such as WiFi-CSI signal processing and deep learning are combined, and theoretical basis and practical experience are finally provided for the application of the deep learning technology in the field of evaluation and identification of passive sensing abnormal behaviors.
The invention is based on the principle of sensing environmental behaviors by WiFi-CSI signals, realizes environmental sensing by utilizing a passive sensing technology and a common wireless router, realizes passive personnel behavior detection by analyzing WiFi-CSI data, and establishes a cross-layer intelligent protection system considering user behaviors, wherein the cross-layer intelligent protection system comprises a function of identifying the position, posture and action of a person and other environmental characteristics. According to the method, the low-level feature representation of the initial data is converted into the high-level abstract feature by using a deep learning algorithm, so that rules hidden in the data are found by using label information, a complex classification learning task is further completed, and the aim of accurately identifying abnormal human behaviors in the intelligent home environment is fulfilled. The abnormal behavior monitoring system designed by the invention provides a method for judging and identifying abnormal behaviors based on WiFi signal perception, which aims to solve the problems of poor privacy protection and low safety and high cost and expense brought by the poor privacy protection and the low safety in the existing human target behavior information identification process.
The technical scheme of the invention is as follows:
an Internet of things intelligent protection system based on WiFi signal perception comprises WiFi signal sending equipment, a WiFi signal receiver, a GUI (Graphical User Interface) control end and a database.
The WiFi signal sending equipment and the WiFi signal receiver are both provided with wireless network cards and are arranged in a room at intervals; the WiFi signal sending equipment is used for sending the WiFi signals with the CSI information, and the WiFi signal receiver is used for acquiring and storing the WiFi-CSI information in the environment.
The GUI control end is a desktop application end, firstly preprocesses data captured by the WiFi-CSI signal receiver, judges and identifies abnormal behaviors through a trained residual error network ResNet50, communicates with a database, stores monitoring behaviors, early warning data and corresponding measures to the database for recording monitoring logs, and monitors and early warns the abnormal behaviors of the user in real time.
An abnormal behavior monitoring method based on WiFi signal perception and deep learning comprises the following specific steps:
the method comprises the following steps: the WiFi signal sending equipment is responsible for sending WiFi-CSI data, and the WiFi signal receiver collects the WiFi-CSI data in the empty environment.
Step two: participants W of abnormal behavior monitoring task n ＝{w 1 ,w 2 ,...,w N And entering a monitoring environment, and making corresponding action according to the requirement of a monitoring task to disturb the WiFi channel state information, and storing the collected WiFi-CSI data by the WiFi signal receiver at the moment.
Step three: and (3) preprocessing the data of the WiFi-CSI collected in the second step by the monitor, then establishing a model, namely a residual error network ResNet50, by using a deep learning algorithm, inputting the preprocessed data into a residual error network ResNet50 for training, and storing network parameters after the training is finished.
Step four: a monitor sends a series of monitoring task requirements by using a GUI control end, at the moment, WiFi signal sending equipment sends WiFi-CSI data to a WiFi signal receiver, a background of the GUI control end preprocesses the data of the WiFi-CSI information and identifies the current behavior by using a trained residual error network ResNet50, whether the behavior is abnormal or not is judged according to an abnormal behavior list preset by the background, and meanwhile, detected behavior information and response measures are output and stored in a database for backup.
The data preprocessing of the WiFi-CSI information in the third step comprises the following specific steps:
(1) spatial transformation and phase transformation. In order to distinguish different channel states by phases of the finally obtained WiFi-CSI data, in a preprocessing stage, an original random phase in the collected WiFi-CSI data is processed mathematically according to a phase transformation principle to eliminate offset of the original random phase, so that phase information with obvious regularity is obtained.
(2) Low-pass filtering and denoising. Discrete Wavelet Transform-DWT (Discrete Wavelet Transform-Discrete Wavelet Transform) is selected, time domain characteristics and frequency domain characteristics are reserved, meanwhile, extracted waveform data volume is compressed, and calculation cost which is paid in the process of directly using the extracted key waveform shape as the characteristics of the key to establish the classifier is reduced. And carrying out Principal Component Analysis (PCA) processing on the filtered signal to achieve the effect of reducing dimensions and denoising. The invention adopts wavelet transform threshold algorithm, and the denoising process is as follows: decomposing the original signal by reasonable wavelet basis and decomposition layer number, performing threshold value processing on the decomposition result, and finally performing wavelet reconstruction to obtain a processed signal.
(3) And (4) data segmentation and feature extraction. And segmenting the denoised WiFi-CSI data, intercepting data of different behaviors, extracting data characteristics of each behavior, and finally generating a sparse matrix.
(4) The data set is divided into a training set, a validation set, and a prediction set. The training set and the validation set are sent to a residual error network ResNet50 for training and a prediction set is used for detecting the training effect of the network.
The deep learning algorithm model in the third step is constructed by the following steps:
(1) a basic ResNet residual block is constructed. There are four types of residual blocks in total, and the number of convolution kernels included in the 3 × 3 convolution layer of the four types of residual blocks increases in order, and is 64, 128, 256, and 512, respectively. The first residual block adopts two convolution layers with the same channel number, so that an input characteristic matrix and an output matrix are directly added and then are sent to an activation function layer; the other three residual blocks are slightly different in design and mainly embodied in the mode of residual connection, and the input characteristic matrix is not directly added to the output matrix like the first residual block, but is added to the output matrix after being subjected to downsampling processing and then is sent to the activation function layer.
(2) The internal structure of the ResNet50 is designed. ResNet50 sets the number of the four residual blocks to 3, 4, 6, 3, the input feature matrix passes through a convolution layer and is normalized, then the input residual block set is calculated in sequence, and finally the classification result is output through dimension conversion of the full-continuous layer.
(3) And adding an activation function layer, a pooling layer, a batch standardization layer, an activation function layer and a tail full-connection layer to ensure efficient training and correct output of the residual error network.
(4) And (4) carrying out feature classification on the output result of the full connection layer by using a Softmax classifier, calculating the probability of each classification, and outputting a prediction label according to the probability (selecting the one with the maximum probability).
(5) Initializing a residual error network, selecting a Nadam optimizer, setting parameters such as learning rate and the like, training the residual error network ResNet50 by using the collected data set, and storing the trained network parameters for identifying user behaviors.
The GUI control end development process in the fourth step is as follows:
(1) firstly, the appearance of an interactive interface is designed according to requirements, and functions to be realized are determined. The abnormal behavior monitoring system of the invention comprises two interfaces: a monitoring interface and a logging interface. The monitoring interface is mainly divided into three parts, namely a WiFi-CSI signal visual dynamic image area, an indoor real environment monitoring area and a monitored behavior display list. The WiFi-CSI signal visual dynamic image is used for observing whether the WiFi-CSI signal normally changes in the whole monitoring process, the indoor real environment monitoring area is used for evaluating whether the system correctly identifies human behaviors according to the collected WiFi-CSI signal, and the behavior display list is used for displaying the monitored behaviors and whether the behaviors belong to abnormal behaviors. In addition, the interface is also provided with an alarm triggering button, and when abnormal behaviors are monitored, the system can not only give corresponding measures, but also trigger an alarm if necessary. The log interface needs to display all behavior information and processing measures in a given time range.
(2) And compiling codes according to the designed interface to realize functions of all parts. The GUI control end of the invention is realized by PyQt5, and the response speed is accelerated by using technologies such as multithread management and the like, so that the user experience is improved.
(3) And configuring the local database. The GUI control end needs to store the monitored behavior information and corresponding measures in a database, and by setting a query time span on a log interface, a background of the GUI control end can screen out all behavior information of which the monitoring time is within the time span from the database and output the behavior information to the interface for display.
(4) And (6) performing software testing. Through testing, the GUI control terminal can successfully identify various behaviors by utilizing a residual error network ResNet50 of a background according to the collected WiFi-CSI data, and outputs and displays behavior information and response measures.
The invention has the beneficial effects that:
compared with other human target behavior identification methods, the method utilizes the WiFi sensing technology to monitor the human activity information, namely the safety of the intelligent home is guaranteed, the detection precision of the system and the accuracy of human target specific behavior identification can be guaranteed, and the defects that the traditional monitoring system is poor in privacy and needs special equipment are overcome. The invention realizes the control of the behavior of the intelligent protection system at the terminal by developing an abnormal behavior monitoring system interface, and puts the behavior and the state of the user into the consideration range of the safety system through the WiFi sensing technology, thereby reducing the use threshold, having flexible strategy and being suitable for the highly dynamic intelligent home system; the established constraint strategy of the application program is considered, the real-time behavior of the user is taken into consideration, the states of the application program and the user behavior are comprehensively considered to formulate an appropriate safety strategy, the problem that the intelligent home system is not humanized is solved, and the participant is encouraged to use the intelligent home better.
Drawings
Fig. 1 is a flow diagram of abnormal behavior monitoring according to the present invention.
Fig. 2 is a flow chart of abnormal behavior monitoring of the present invention.
Fig. 3 is a schematic diagram of the WiFi signal acquisition and preprocessing process.
Fig. 4 is a flow chart of a deep learning abnormal behavior recognition algorithm.
Fig. 5 is a network configuration diagram of the residual network ResNet 50.
Fig. 6 is a residual block diagram included in the residual network ResNet 50.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
Fig. 1 is a view of an abnormal behavior monitoring process scenario according to the present invention, and an intelligent protection system according to the present invention is composed of a WiFi signal transmission device, a WiFi signal receiver, a participant, a monitor, a GUI control terminal, and a database, where the monitor controls the WiFi signal receiver through the GUI control terminal, the participant generates a series of actions, and at this time, the WiFi signal receiver obtains WiFi-CSI data. After being preprocessed, the data are sent to a residual error network ResNet50 to be recognized, whether the behavior is abnormal is judged according to an abnormal behavior list preset in a background, and meanwhile, detected behavior information and corresponding measures are output and stored in a database for backup.
Fig. 2 is a timing diagram of an abnormal behavior monitoring process according to the present invention, and the present invention provides an abnormal behavior monitoring process based on WiFi signal sensing, which specifically includes the following steps:
the method comprises the following steps: the WiFi signal sending equipment is responsible for sending WiFi-CSI data, and the WiFi signal receiver collects the WiFi-CSI data in the empty environment.
Step two: participants W of abnormal behavior monitoring task n ＝{w 1 ,w 2 ,...,w N And entering a monitoring environment, and making corresponding action according to the requirement of a monitoring task to disturb the WiFi channel state information, and storing the collected WiFi-CSI data by the WiFi signal receiver at the moment.
Step three: and (3) preprocessing the data of the WiFi-CSI collected in the second step by the monitor, then establishing a model, namely a residual error network ResNet50, by using a deep learning algorithm, inputting the preprocessed data into the residual error network for training, and storing the network parameters after the training is finished.
Step four: a monitor sends a series of monitoring task requirements by using a GUI control end, at the moment, WiFi signal sending equipment sends WiFi-CSI data to a WiFi signal receiver, a background of the GUI control end preprocesses the data of the WiFi-CSI information and identifies the current behavior by using a trained residual error network ResNet50, whether the behavior is abnormal or not is judged according to an abnormal behavior list preset by the background, and meanwhile, detected behavior information and response measures are output and stored in a database for backup.
Fig. 3 is a schematic view of a WiFi-CSI data acquisition and preprocessing process, wherein the data preprocessing specifically includes the following steps:
(1) spatial transformation and phase transformation. In order to distinguish different channel states by phases of the finally obtained WiFi-CSI data, in a preprocessing stage, an original random phase in the collected WiFi-CSI data is processed mathematically according to a phase transformation principle to eliminate offset of the original random phase, so that phase information with obvious regularity is obtained.
(2) Low-pass filtering and denoising. Discrete Wavelet Transform-DWT (Discrete Wavelet Transform-Discrete Wavelet Transform) is selected, time domain characteristics and frequency domain characteristics are reserved, meanwhile, extracted waveform data volume is compressed, and calculation cost which is paid in the process of directly using the extracted key waveform shape as the characteristics of the key to establish the classifier is reduced. And carrying out Principal Component Analysis (PCA) processing on the filtered signal to achieve the effect of reducing dimensions and denoising. In this embodiment, a wavelet transform threshold algorithm is adopted, and the denoising process is as follows: decomposing the original signal by reasonable wavelet base and decomposition layer number, carrying out threshold value processing on the decomposition result, and finally carrying out wavelet reconstruction to obtain the processed signal.
(3) And (4) data segmentation and feature extraction. And segmenting the denoised WiFi-CSI data, intercepting data of different behaviors, extracting data characteristics of each behavior, and finally generating a sparse matrix.
(4) The data set is divided into a training set, a validation set, and a prediction set. The training set and the validation set are sent to a residual error network ResNet50 for training and a prediction set is used for detecting the training effect of the network.
Before the detailed steps of establishing the deep learning algorithm model of the invention are explained, the following description is made to the problems:
firstly, various abnormal behavior data in the intelligent home environment are collected through experiments and are sent to a neural network for training and prediction, the recognition capability of the neural network on CSI signals is analyzed, and the structure and parameters of the neural network are continuously adjusted according to the experiment results, so that the feasibility of the method is improved. Finally, the invention determines to adopt a depth residual error network ResNet50 as a main structure, and optimizes parameters to improve the identification accuracy. And sending the collected various abnormal behavior data to a neural network by using a server for training and predicting, and storing the trained network parameters into a file.
Fig. 4 is a flowchart of an abnormal behavior recognition algorithm adopted by the present invention, in which a residual error network ResNet50 is adopted for recognition, and a process for constructing a residual error network ResNet50 specifically includes the following steps:
(1) a basic ResNet residual block is constructed. There are four types of residual blocks in total, and the number of convolution kernels included in the 3 × 3 convolution layer of the four types of residual blocks increases in order, and is 64, 128, 256, and 512, respectively. The first residual block adopts two convolution layers with the same channel number, so that an input characteristic matrix and an output matrix are directly added and then are sent to an activation function layer; the other three residual blocks are slightly different in design and mainly embodied in the mode of residual connection, and the input characteristic matrix is not directly added to the output matrix like the first residual block, but is added to the output matrix after being subjected to downsampling processing and then is sent to the activation function layer.
(2) The internal structure of the ResNet50 is designed. ResNet50 sets the number of the four residual blocks to 3, 4, 6, 3, the input feature matrix passes through a convolution layer and is normalized, then the input residual block set is calculated in sequence, and finally the classification result is output through dimension conversion of the full-continuous layer.
(3) And adding an activation function layer, a pooling layer, a batch standardization layer, an activation function layer and a tail full-connection layer to ensure efficient training and correct output of the residual error network. In this embodiment, the number of nodes in the full connection layer is set to 6, that is, 6 types of abnormal behaviors are given, which are: the door lock is not closed, the participant is not in bed but the alarm is not closed, the baby is at the window, the participant is in office but the sound is not closed, the television is in open but no one watches for a long time, and the participant is not seated.
(4) And (4) carrying out feature classification on the output result of the full connection layer by using a Softmax classifier, calculating the probability of each classification, and outputting a prediction label according to the probability (selecting the one with the maximum probability).
(5) Initializing a residual network, selecting a Nadam optimizer, setting the learning rate to be 0.005, training the residual network ResNet50 by using the collected data set, and storing the trained network parameters for identifying user behaviors. Trained, the residual network ResNet50 can identify the above six types of abnormal behavior.
Fig. 5 is a network configuration diagram of the residual network ResNet 50. The parameters of each residual block involved in the residual network ResNet50 construction step described above are further explained.
Fig. 6 is a diagram of residual blocks included in the residual network ResNet 50. The internal structure of each residual block involved in the residual network ResNet50 construction step described above is further explained.
The GUI control end for monitoring the abnormal behaviors of the smart home is developed based on the method, the display of the monitoring picture and the visualization of the WiFi-CSI signal in the monitoring process can be realized, the background of the GUI control end realizes the identification of the abnormal behaviors by utilizing the ResNet50 network which is trained, and the identified behavior information and specific response measures are output and displayed. The specific development process of the GUI control end is as follows:
(1) firstly, the appearance of an interactive interface is designed according to requirements, and functions to be realized are determined. The abnormal behavior monitoring system of the invention comprises two interfaces: a monitoring interface and a logging interface. The monitoring interface is mainly divided into three parts, namely a WiFi-CSI signal visual dynamic image area, an indoor real environment monitoring area and a monitored behavior display list. The WiFi-CSI signal visual dynamic image is used for observing whether the WiFi-CSI signal normally changes in the whole monitoring process, the indoor real environment monitoring area is used for evaluating whether the system correctly identifies human behaviors according to the collected WiFi-CSI signal, and the behavior display list is used for displaying the monitored behaviors and whether the behaviors belong to abnormal behaviors. In addition, the interface is also provided with an alarm triggering button, and when abnormal behaviors are monitored, the system can not only give corresponding measures, but also trigger an alarm if necessary. The log interface needs to display all behavior information and processing measures in a given time range.
(2) And compiling codes according to the designed interface to realize functions of all parts. The GUI control end of the invention is realized by PyQt5, and the response speed is accelerated by using technologies such as multithread management and the like, so that the user experience is improved.
(3) And configuring the local database. The GUI control end needs to store the monitored behavior information and corresponding measures in a database, and by setting a query time span on a log interface, a background of the GUI control end can screen out all behavior information of which the monitoring time is within the time span from the database and output the behavior information to the interface for display.
(4) And (6) performing software testing. Through testing, the GUI control terminal can successfully identify various behaviors by utilizing a residual error network ResNet50 of a background according to the collected WiFi-CSI data, and outputs and displays behavior information and response measures.
Claims (3)
1. A method for monitoring abnormal behaviors of an Internet of things intelligent protection system based on WiFi signal sensing is characterized in that the system comprises WiFi signal sending equipment, a WiFi signal receiver, a GUI control end and a database;
the WiFi signal sending equipment and the WiFi signal receiver are both provided with a wireless network card, and the WiFi signal sending equipment and the WiFi signal receiver are arranged in a room at intervals; the WiFi signal sending equipment is used for sending WiFi signals with CSI information, and the WiFi signal receiver is used for acquiring and storing WiFi-CSI information in the environment;
the GUI control end is a desktop application end, preprocesses data captured by the WiFi-CSI signal receiver, judges and identifies abnormal behaviors through a trained residual error network ResNet50, communicates with a database, stores monitoring behaviors, early warning data and corresponding measures into the database for recording monitoring logs, and monitors and early warns users of abnormal behaviors in real time;
the method comprises the following steps:
the method comprises the following steps: the WiFi signal sending equipment is responsible for sending WiFi-CSI data, and the WiFi signal receiver collects the WiFi-CSI data in the empty environment;
step two: participants W of abnormal behavior monitoring task n ＝{w 1 ,w 2 ,...,w N Entering a monitoring environment, and making corresponding action according to the requirement of a monitoring task to disturb the WiFi channel state information, wherein the WiFi signal receiver stores the collected WiFi-CSI data;
step three: the monitor preprocesses the data of the WiFi-CSI data collected in the second step, then establishes a residual error network ResNet50 by using a deep learning algorithm, inputs the preprocessed data into a residual error network ResNet50 for training, and stores network parameters after the training is finished;
step four: a monitor sends a series of monitoring task requirements by using a GUI control end, at the moment, WiFi signal sending equipment sends WiFi-CSI data to a WiFi signal receiver, a background of the GUI control end preprocesses the data of the WiFi-CSI information and identifies the current behavior by using a trained residual error network ResNet50, whether the behavior is abnormal or not is judged according to an abnormal behavior list preset by the background, and meanwhile, detected behavior information and response measures are output and stored in a database for backup;
in the third step, the deep learning algorithm model is constructed by the following steps:
(1) constructing basic ResNet residual block
The residual blocks are totally of four types, the number of convolution kernels contained in the 3 x 3 convolution layer in the four types of residual blocks is sequentially increased and is respectively 64, 128, 256 and 512; the first residual block adopts two convolution layers with the same channel number, so that an input characteristic matrix and an output matrix are directly added and then are sent to an activation function layer; the input characteristic matrixes of other three residual blocks are subjected to downsampling processing and then added with the output matrix to be sent to an activation function layer;
(2) internal structure of ResNet50
ResNet50 sets the number of the four residual blocks as 3, 4, 6 and 3, inputs the characteristic matrix to pass through a convolution layer and carry out normalization processing, inputs the residual block set to calculate in sequence, and finally carries out dimension conversion through a full-connection layer to output classification results;
(3) adding an activation function layer, a pooling layer, a batch standardization layer, an activation function layer and a tail full-connection layer to ensure efficient training and correct output of a residual error network;
(4) performing feature classification on the output result of the full connection layer by using a Softmax classifier, calculating the probability of each classification, and selecting the one with the maximum probability to output a prediction label;
(5) initializing a residual error network, selecting a Nadam optimizer, setting a learning rate, training the residual error network ResNet50 by using the collected data set, and storing the trained network parameters for identifying user behaviors.
2. The method of claim 1, wherein in the third step, the data preprocessing of WiFi-CSI data is as follows:
(1) spatial transformation and phase transformation
Firstly, performing mathematical processing on an original random phase in collected WiFi-CSI data according to a phase transformation principle to eliminate the offset of the original random phase and obtain phase information with regularity;
(2) low pass filtering and denoising
Discrete wavelet transform is selected, time domain characteristics and frequency domain characteristics are reserved, meanwhile, extracted waveform data amount is compressed, and calculation cost which is paid in the process of directly using the extracted key waveform shape as the characteristics of the key to establish the classifier is reduced; principal component analysis processing is carried out on the filtered signals, and the effect of reducing dimensions and denoising is achieved; adopting wavelet transform threshold algorithm, the denoising process is as follows: decomposing an original signal through a wavelet basis and the number of decomposition layers, performing threshold value processing on a decomposition result, and performing wavelet reconstruction to obtain a processed signal;
(3) data segmentation and feature extraction
Segmenting the denoised WiFi-CSI data, intercepting data of different behaviors, extracting data characteristics of each behavior, and finally generating a sparse matrix;
(4) partitioning a data set into a training set, a validation set, and a prediction set
The training set and the validation set are sent to a residual error network ResNet50 for training and a prediction set is used for detecting the training effect of the network.
3. The method according to claim 1 or 2, wherein in the fourth step, the GUI control end development process is as follows:
(1) designing the appearance of the interactive interface according to requirements, and determining functions to be realized
The GUI control includes two interfaces: a monitoring interface and a log interface;
the monitoring interface is divided into a WiFi-CSI signal visual dynamic image area, an indoor real environment monitoring area and a monitored behavior display list; the WiFi-CSI signal visual dynamic image is used for observing whether the WiFi-CSI signal normally changes in the whole monitoring process; the indoor real environment monitoring area is used for evaluating whether the system correctly identifies human behaviors according to the acquired WiFi-CSI signals; the behavior display list is used for displaying the monitored behaviors and whether the behaviors belong to abnormal behaviors; in addition, the interface is also provided with an alarm triggering button, when abnormal behaviors are monitored, the system can not only give corresponding measures, but also trigger an alarm if necessary;
the log interface needs to display all behavior information and processing measures in a given time range;
(2) the function of the interface on the GUI control end is realized by utilizing PyQt 5;
(3) configuring a local database
The GUI control end needs to store the monitored behavior information and corresponding countermeasures into a database, and the background of the control end screens out all behavior information of which the monitoring time is within the time span from the database and outputs the behavior information to an interface for display by setting a query time span on a log interface;
(4) software testing
Through testing, the GUI control terminal can successfully identify various behaviors by utilizing a residual error network ResNet50 of a background according to the collected WiFi-CSI data, and outputs and displays behavior information and response measures.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111183088.1A CN113965362B (en) | 2021-10-11 | 2021-10-11 | Internet of things intelligent protection system and method based on WiFi signal sensing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111183088.1A CN113965362B (en) | 2021-10-11 | 2021-10-11 | Internet of things intelligent protection system and method based on WiFi signal sensing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113965362A CN113965362A (en) | 2022-01-21 |
| CN113965362B true CN113965362B (en) | 2022-08-09 |
Family
ID=79464039
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111183088.1A Active CN113965362B (en) | 2021-10-11 | 2021-10-11 | Internet of things intelligent protection system and method based on WiFi signal sensing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113965362B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108901021A (en) * | 2018-05-31 | 2018-11-27 | 大连理工大学 | A kind of deep learning identification system and method based on channel state information of wireless network |
| CN111597877A (en) * | 2020-04-02 | 2020-08-28 | 浙江工业大学 | Fall detection method based on wireless signals |
| CN113033654A (en) * | 2021-03-23 | 2021-06-25 | 云南大学 | Indoor intrusion detection method and system based on WiFi channel state information |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10375585B2 (en) * | 2017-07-06 | 2019-08-06 | Futurwei Technologies, Inc. | System and method for deep learning and wireless network optimization using deep learning |
| CN111460891B (en) * | 2020-03-01 | 2023-05-26 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Automatic driving-oriented vehicle-road cooperative pedestrian re-identification method and system |
| CN113015093B (en) * | 2021-01-29 | 2024-04-05 | 辽宁大学 | Indoor wireless positioning method based on three-dimensional depth residual error neural network |
-
2021
- 2021-10-11 CN CN202111183088.1A patent/CN113965362B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108901021A (en) * | 2018-05-31 | 2018-11-27 | 大连理工大学 | A kind of deep learning identification system and method based on channel state information of wireless network |
| CN111597877A (en) * | 2020-04-02 | 2020-08-28 | 浙江工业大学 | Fall detection method based on wireless signals |
| CN113033654A (en) * | 2021-03-23 | 2021-06-25 | 云南大学 | Indoor intrusion detection method and system based on WiFi channel state information |
Non-Patent Citations (2)
| Title |
|---|
| 基于WiFi信号的人体动作识别系统;肖玲等;《北京邮电大学学报》;20171124(第03期);第1-5页 * |
| 基于多天线FMCW雷达的人体行为识别方法;田增山等;《重庆邮电大学学报(自然科学版)》;20201015(第05期);第1-8页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113965362A (en) | 2022-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110428522A (en) | A kind of intelligent safety and defence system of wisdom new city | |
| CN110363131B (en) | Abnormal behavior detection method, system and medium based on human skeleton | |
| CN109154976A (en) | Pass through the system and method for machine learning training object classifier | |
| KR102396888B1 (en) | Real-time smart fire fighting sensing and monitoring system | |
| CN106503666A (en) | A kind of method for safety monitoring, device and electronic equipment | |
| US10165228B2 (en) | Sensor event assessor training and integration | |
| CN109815904A (en) | Fire identification method based on convolutional neural network | |
| CN109033973A (en) | Monitoring and early warning method, apparatus and electronic equipment | |
| CN110717461A (en) | Fatigue state identification method, device and equipment | |
| CN113076903A (en) | Target behavior detection method and system, computer equipment and machine readable medium | |
| Wang et al. | Online fault-tolerant dynamic event region detection in sensor networks via trust model | |
| CN101316371A (en) | Flame detecting method and device | |
| CN110022466A (en) | A kind of video analysis platform and its control method based on wisdom big data | |
| CN101790717B (en) | Machine vision system for enterprise management | |
| CN113965362B (en) | Internet of things intelligent protection system and method based on WiFi signal sensing | |
| CN111798356B (en) | Rail transit passenger flow abnormal pattern recognition method based on big data | |
| CN111753587A (en) | Method and device for detecting falling to ground | |
| CN109871456B (en) | Method and device for analyzing relationship between watchmen and electronic equipment | |
| CN105451235A (en) | Wireless sensor network intrusion detection method based on background updating | |
| CN116248346A (en) | Smart city-oriented CPS network security situation awareness establishing method and system | |
| CN106921846A (en) | Video mobile terminal legacy detection means | |
| CN115909144A (en) | Method and system for detecting abnormity of surveillance video based on counterstudy | |
| CN112601054B (en) | Pickup picture acquisition method and device, storage medium and electronic equipment | |
| Cai et al. | Interactive visualization of network anomalous events | |
| ISMAIL | DEVELOPMENT OF ENHANCED BAYESIAN MODEL FOR DETECTION OF COVERT MEMBERS IN CRIMINAL NETWORKS USING TELECOMMUNICATION METADATA |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |